Combinatorial Optimization of Reconfigurable Intelligence Surfaces at Wireless Endpoints using the Ising Hamiltonian Model

Ross, Charles; Lim, Qijian; You, Minglei; Gradoni, Gabriele; Peng, Zhen

Abstract

The reconfigurable intelligent surface (RIS) based on discrete meta-surfaces with tunable elements has been widely studied in wireless communication and electromagnetics communities. Researchers have devoted substantial efforts to investigating large-scale optimization algorithms that achieve desired channel conditions. This is particularly challenging in low-complexity RIS architectures with minimal hardware and no sensing capabilities. In this paper, we propose a physics-oriented computational framework that optimizes RIS configuration using feedback (i.e. received power) from the wireless endpoints. The new idea is grounded on the isomorphism between the power of the RIS-aided channel transfer function and the Hamiltonian of Ising spin glass model. The problem of optimizing RIS configuration is converted into finding the ground state of the Ising Hamiltonian. The coefficients of the Ising spin bias and interactions are learned onsite by a generic supervised learning model known as a factorization machine, which enables the possibility of fast optimization adapting to dynamic wireless environments. The performance of the proposed work is demonstrated in some representative wireless propagation scenarios.

Citation

Ross, C., Lim, Q., You, M., Gradoni, G., & Peng, Z. (2023, July). Combinatorial Optimization of Reconfigurable Intelligence Surfaces at Wireless Endpoints using the Ising Hamiltonian Model. Presented at IEEE International Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting (2023), Portland, Oregon, USA

• Presentation Conference Type: Presentation / Talk
• Conference Name: IEEE International Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting (2023)
• Start Date: Jul 23, 2023
• End Date: Jul 28, 2023
• Deposit Date: Aug 2, 2023
• Public URL: https://nottingham-repository.worktribe.com/output/23732504
• Publisher URL: https://2023.apsursi.org/view_paper.php?PaperNum=2160